New Adaptive Perceptual Quantization Method for HDR Video Coding
Hangfei She, Yang Song, Mei Yu, Gangyi Jiang · 2021
High dynamic range (HDR) videos need to be perceptually quantized into 10bit integer for compression, this perceptually quantization (PQ) process may lead to distortion and coding efficiency decreasing. To solve this problem, this paper proposes a new adaptive PQ (APQ) method. Firstly, the contrast distortion caused by PQ is analyzed, and the regions prone to contrast distortion are extracted. Then, by exploring the distribution characteristics in each channel, the pixel probability densities in different luminance ranges are counted and the distribution of the quantized 10bit HDR video is optimized by combining with visual perception. Furthermore, the regions prone to contrast distortion is optimized to protect the contrast information. Finally, the mapping for original HDR video signal is obtained by using the optimization strategy of 10-bit quantized data combined with inverse PQ. The experimental results show that the proposed APQ method can decrease the BD-rate in an average of 8.73 with HDR-VDP-2 as quality metric, which proves that the proposed APQ method can effectively reduce the distortion and improve coding performance.